Detecting Thai Messages Leading to Deception on Facebook

摘要

Social network has become a very popular communication for Thai people, especially Facebook. Unfortunately, this popularity also attracts deceiver spreading malicious messages to other users. Some messages lead to deception. This paper studies Thai messages posted on Facebook that lead to deception. We try to investigate different approaches to detect deceptive messages and find dominant words. To detect deceptive messages, the dataset is retrieved from Facebook pages. Next, content-based and context-based features are extracted from the dataset. Two algorithms, i.e. SVM and KNN, are applied to perform a prediction. We construct the experiments to investigate context-based and content-based features for detecting deceptive messages. The experimental results show that the context-based features gives the best performance and the F-measure for predicting deceptive messages achieves 99 % when using SVM classifier. In addition, dominant words in deceptive messages and truthful messages are reported in our work.